Out of breath when talking

Searched over 200M research papers

Research Analysis by Consensus

Beta · 19 papers

Consensus Meter

Yes

Possibly

Understanding the Phenomenon of Being Out of Breath When Talking

Vocal Tract Characteristics in Out-of-Breath Speech

When a person is out of breath, particularly after physical exercise, their vocal tract undergoes noticeable changes. Research indicates that the first four formants of the vocal tract show a lowering in their average frequency. Additionally, the bandwidths of the first two formants (BF1 and BF2) widen, while the other two formants narrow slightly. These changes are not uniform across all formants for every speaker, suggesting individual variability in how physical exertion affects speech production.

Fourier Model Analysis for Out-of-Breath Speech

A novel method using Fourier model-based features has been developed to analyze out-of-breath speech. This method evaluates the differences and ratios of Fourier parameters, such as amplitude and frequency, between contiguous values. The analysis has shown that these features can effectively differentiate between normal speech and out-of-breath speech, as well as distinguish varying levels of breath emission. The proposed features outperform traditional features like breathiness, mel frequency cepstral coefficients (MFCC), and Teager energy operator-based features in classification tasks.

Adaptive Wavelet Transform for Breathiness Detection

The breathing process significantly impacts speech production, especially after heavy physical exercise. An adaptive wavelet approach using Empirical Wavelet Transform (EWT) has been employed to analyze the subband characteristics of out-of-breath speech. This method produces intrinsic mode signals from which features are extracted and combined using Temporal Pyramid Matching (TPM). The separability of these features is evaluated using Support Vector Machine (SVM) and Artificial Neural Network (ANN) classifiers, with SVM showing a higher accuracy of 75.41% in distinguishing normal from out-of-breath speech.

Video-Based Breath Signal Extraction

A cost-effective method for extracting breath signals using video devices like mobile phones and computers has been developed. This method captures breath signals during both neutral and post-exercise conditions. Findings indicate that out-of-breath conditions result in stronger inhalation and exhalation phases and a shorter average breath cycle duration, primarily due to a reduced exhalation phase. Combining these breath features with MFCC baseline features improves the performance of classifiers like SVM and logistic regression, achieving an unweighted average recall and F1-score of approximately 70%.

Breathing Cues in Conversation

Breathing cues play a crucial role in managing speaking turns during conversations. A new categorization of turn-taking events considers whether the original speaker inhales before continuing to speak. This criterion can serve as a proxy for the pragmatic completeness of the previous utterance and the interruptive nature of incoming speech. Additionally, breath holds are often used in response to incoming talk rather than as a turn-holding cue. These findings suggest that breathing signals can uncover hidden turn-taking events that are not apparent in silence-based interaction representations.

Conclusion

The phenomenon of being out of breath when talking involves complex changes in vocal tract characteristics, breath signal patterns, and conversational dynamics. Advanced methods like Fourier model analysis, adaptive wavelet transforms, and video-based breath signal extraction provide valuable insights into these changes. Understanding these aspects can improve the classification and analysis of out-of-breath speech, offering potential applications in various fields such as speech therapy, sports science, and human-computer interaction.